This article was downloaded by: [Peter Henne] On: 06 December 2011, At: 07:50 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Terrorism and Political Violence Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ftpv20 The Ancient Fire: Religion and Suicide Terrorism a Peter S. Henne a Department of Government, Georgetown University, Washington, DC, USA Available online: 06 Dec 2011 To cite this article: Peter S. Henne (2012): The Ancient Fire: Religion and Suicide Terrorism, Terrorism and Political Violence, 24:1, 38-60 To link to this article: http://dx.doi.org/10.1080/09546553.2011.608817 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. Terrorism and Political Violence, 24:38–60, 2012 Copyright # Taylor & Francis Group, LLC ISSN: 0954-6553 print=1556-1836 online DOI: 10.1080/09546553.2011.608817 The Ancient Fire: Religion and Suicide Terrorism PETER S. HENNE Department of Government, Georgetown University, Washington, DC, USA Does religion lead to greater destructiveness from suicide terrorism? And if so, how does it influence this form of political violence? Recent analyses of terrorism point to the significance of religion, but are divided as to whether religion itself matters, or certain types of religious terrorist groups are actually driving suicide terrorist viol- ence. This article draws on social movement theory and recent work in the study of suicide terrorism to argue that religion influences the severity of suicide terrorist attacks as an ideology groups use to justify their struggle and gain public support. This effect occurs regardless of a group’s goals or organizational nature. The theory is tested using a generalized estimating equation to account for multiple attacks by several groups. The study finds that the religious ideology of a group greatly increases the number of deaths from a suicide attack, even if varying group motiva- tions and structural factors are taken into account. The article helps to clarify the effect of religion on contemporary terrorism, contributing to the study of both terrorism and religion and politics. Keywords Islam, quantitative methods, religion, social movement theory, terrorism Must the ancient fire not some day flare up much more terribly, after much longer preparation? 1 —Friedrich Nietzsche In November 2005, a series of suicide bombings in Amman, Jordan killed approxi- mately 60 people and wounded over 100. The attacks—directed by Abu Musab al-Zarqawi’s al-Qaeda in Iraq—were characterized by their coordination, severity, and their perpetrator’s desire to institute Islamic rule in the region. Exceptionally Peter S. Henne is a doctoral candidate in the Department of Government at Georgetown University. Earlier versions of this article were presented at the 2009 Midwest Political Science Association Conference, and at a 2010 meeting of the Georgetown University International Theory and Research Seminar. The author would like to thank Andrew Bennett, David Buckley, Daniel Byman, Erica Chenoweth, David Edelstein, Daniel Hopkins, Philip Kapusta, Matthew Kroenig, Tristan Mabry, Elizabeth Mercurio, Daniel Nexon, Christopher Preble, Zacchary Ritter, Erik Voeten, James Vreeland, and the editors and reviewers of Terrorism and Political Violence. All remaining errors are the sole responsibility of the author. Address correspondence to Peter S. Henne, Department of Government, Georgetown University, Box 571034, Intercultural Center 681, Washington, DC 20057, USA. E-mail: The Ancient Fire: Religion and Suicide Terrorism 39 violent suicide attacks like these have been increasing over the past few decades, and policymakers and scholars have been scrambling to explain this disturbing phenom- enon, and what connection it has to the resurgence of religion in the last quarter of the twentieth century. Many early explanations of suicide terrorism argued that the nature of Islam or conditions in the ‘‘Muslim world’’ were the cause of this destructive 2 phenomenon. Later analyses, however, pointed to non-religious factors such as occupation, governmental repression, ethnic fractionalization, or inter-group com- 3 petition. More recently, there has been a progressive development in the debate over religion and terrorism, with many scholars accepting that religion does influence political violence, and focusing on the means through which religion affects violence. Some admit that religion helps to motivate combatants, but its significance is sec- 4 ondary to non-religious factors. Others point to organizational and motivational 5 characteristics of groups that—along with religion—lead to severe violence. This development in the study of religion and political violence has led to diverging opinions among scholars of suicide terrorism, specifically concerning how religion influences suicide terrorism. This takes on a particular salience when looking at the destructiveness of groups’ suicide attacks; groups may adopt suicide terrorism for a variety of reasons—many of which have nothing to do with religi- on—but religion-related beliefs and motivations can still affect how many fatalities their attacks cause. It may be that fundamentalist groups focused solely on advanc- ing their version of religion on a global stage cause the most destruction from suicide 6 attacks. Or the most dramatic violence may arise when groups with religious moti- 7 vations are fighting in the context of a nationalist struggle. Alternately, nationalism, repression, or economic deprivation may be more important than religion in deter- mining the severity of suicide terrorism. This article builds on studies that point to the unique effects of religious beliefs on political violence by applying their insights to variations in the severity of suicide terrorist attacks. It also helps to clarify the mechanism through which religious ideol- ogy influences violence by drawing on social movement theory work on framing. It argues that group motivations and organizational characteristics affect group beha- vior, but do not explain the level of violence among suicide attacks. When a religious ideology is utilized in instances of political violence, it can lead to combatants per- ceiving their struggle as a sacred one, and becoming disassociated from local factors. Yet this is not dependent on specific group motivations; instead, all religious violence, whether ethnoreligious or fundamentalist, is more destructive than violence driven by non-religious motivations. Once a group engages in suicide attacks, the religious framing of their actions can lead to an emphasis on causing numerous fatal- ities, whatever the group’s targets or aspirations. While group characteristics and goals are undoubtedly important, the nature of religious ideology itself is what explains the severity of suicide terrorism. This article tests the role of religious ideology in suicide terrorist violence through a quantitative analysis of suicide terrorist attacks over the past three dec- ades. Using a generalized estimating equation (GEE), the study demonstrates that suicide attacks by groups with a religious ideology are more violent than those with nationalist or leftist ideologies; this finding holds up even if the effect of both funda- mentalist and ethnoreligious groups are accounted for. Also, this level of violence cannot be attributed to socioeconomic factors or the demographic makeup of the group. Downloaded by [Peter Henne] at 07:50 06 December 2011 40 P. S. Henne The article makes several contributions to relevant theoretical literatures. First, it clarifies and corroborates recent findings in the study of religion and terrorism. Religious ideology, rather than groups’ goals, is the most important factor in a group’s level of suicide terrorism. Groups motivated by a mixture of religious and ethnic grievances may be more likely to launch violent terrorist attacks, while groups trying to advance a ‘‘pure’’ form of religion may be the most brutal; when looking specifically at variations in the number of deaths these groups cause from their attacks, however, these characteristics matter less than the broader effects of religion on violence. Second, by conceptualizing religion’s effects on violence as a group-level ideology that frames terrorist actions, the article helps to specify the mechanisms through which religion influences conflict. The article therefore complements and extends existing studies of religion and political violence. While many existing arguments focus on the causes of suicide terrorism or terrorism in general, their conclusions can extend to the severity of groups’ suicide attacks. If Islamic traditions lend themselves to greater violence, this should be reflected in the damage groups cause through their attacks. Likewise, if the desper- ation caused by political repression or economic deprivation causes groups to resort to suicide terrorism, that same desperation should drive them to maximize their attacks’ impact. Finally, if anger over occupying forces is sufficient to introduce suicide terrorism into a conflict, this should motivate groups to harm the other side as much as possible. A few points on definitions are required. The study uses Bruce Hoffman’s definition of terrorism as ‘‘the deliberate creation and exploitation of fear through 8 violence or the threat of violence in pursuit of political change.’’ Suicide terrorism is—following Pape—an attack in which the attacker does not expect to survive the mission and often employs a method of attack . . . that requires his or her death in 9 order to succeed.’’ An ethnoreligious group is a group motivated by both religious and nationalist concerns, or a group that advances a religious ideology in the context 10 of a territorial dispute. A fundamentalist group is a group that is primarily con- 11 cerned with advancing its interpretation of the religion. A religious group, then, is a terrorist group that advances a religious ideology; this category includes both ethnoreligious and fundamentalist groups and addresses a group’s ideological position, not its demographic makeup. The study proceeds in three parts. First, it presents a theory based on religious ideology and the framing of violent political contention. Next, it discusses the methodology of the analysis. Finally, it presents the findings of the study and their applicability to the theoretical arguments. Religious Ideology and Framing in Suicide Terrorism This article draws on existing works on religion and terrorism, synthesizing them with social movement theory and applying them to the specific area of variation in the level of suicide terrorism violence. It posits that religion affects the level of suicide terrorism violence through its mobilization by a group as an ideology. Some terrorist groups draw on an ideology that defines their struggle in religious terms, justifying dramatic actions in support of their claims and conducting extremely fatal attacks. This ideology functions as a collective action frame, and thus influences the perceptions and tactics of a group regardless of their ultimate goals. oD wnloadedb yeP[ tereH nnea ] t:70 eD 60 05 cember2 10 1 The Ancient Fire: Religion and Suicide Terrorism 41 Religion and Political Violence The study approaches religion as a specific type of belief system, whose infusion into politics can lead to dramatic effects. Religion is a ‘‘system of beliefs and practices oriented towards the sacred or supernatural,’’ rather than one based on ethnic pride 12 or liberal values. Connecting religion and politics leads to dramatic forms of polit- ical expression and the fusion of local disputes with global religious struggles. The political mobilization of religious beliefs challenges states’ legitimacy to a greater extent than would non-religious movements, raising the normative stakes and justi- 13 fying extreme action. Moreover, the universal nature of many religions connects individuals with coreligionists across multiple societies, so religiously-inspired con- tention will be placed in the context of global struggles, rather than influenced solely 14 by one country’s socioeconomic and political conditions. Religious influences on politics thus increase the intensity of groups’ devotion to their goals and connect them to actors and situations beyond their society. When religion’s political expression occurs in the context of terrorism the result can be dramatic acts of violence. In an attempt to make sense of disorder in the world, religious groups will externalize religion’s struggle against evil onto a con- 15 flict. Combatants are not viewed as rivals to power or economic resources, but as personifications of evil and the enemy of their religion itself. Moreover, the ‘‘out group’’ includes anyone who does not follow the terrorists’ particular interpret- ation of a religion, so the violence could be directed against civilians in the same community as the group. This leads to severe, dramatic violence and conflicts that 16 are difficult to resolve due to their sacred underpinnings. Among terrorist groups, this effect of religion translates most directly into the severity of the group’s attacks. As several scholars have noted, terrorist groups that view their struggle as a sacred 17 duty are characterized by indiscriminately violent acts. This can be contrasted with nationalist groups, which—even though they conduct horrific attacks—are usually more restrained than religious groups; nationalists must demonstrate they represent an ethnic group so they are constrained by local standards and public opinion, which can translate into attempts to minimize civilian damage, especially among local 18 communities. This connection of religion to political violence has been particularly dramatic in recent decades among Muslims. The most violent terrorist groups currently operat- ing are for the most part Muslim, and many claim to be acting based on Islamic beliefs. Yet, Muslims are no more likely to be involved in conflict than any other 19 religious groups, and severe violence can be produced by offshoots of any religion. Also, not all Muslims who participate in the activities of terrorist groups are driven by religious sentiment. The prevalence of Muslims in conflicts is thus more likely the 20 result of contemporary structural factors rather than something inherent in Islam. The explanation for the violence perpetrated by self-identified Islamic groups there- fore lies in the dynamics of the social movements they constitute. Religious Ideology and Framing in Suicide Terrorism Many contemporary scholars of suicide terrorism accept the above discussion on religion’s role in political violence, but disagree on how it matters, with many point- ing not to religion per se but specific types of religious motivations. Some claim resistance to occupation or intra-group competition drive suicide terrorism, but Downloaded by [Peter Henne] at 07:50 06 December 2011 42 P. S. Henne 21 religion can be a motivating force for groups. Others argue that religious motiva- tions in the context of ethnic conflict lead groups to commit severe suicide terrorist 22 attacks. Still others posit that fundamentalist groups that see themselves as part of a global struggle to defend and advance their religion will be driven to kill as many of 23 their enemies as possible. While each of these views has merit, there are problems when applying them to the issue of variations in suicide terrorism violence. Groups may adopt suicide ter- rorism because of non-religious factors, but if their religious beliefs affect how they conduct attacks, then ignoring these beliefs leaves most of the phenomenon unexplained. Explanations focusing on varying group motivations do point to the mechanisms connecting religious beliefs to suicide terrorism, but are still limited. While group motivations are important, specific attacks are often determined by rational calculations or pragmatic considerations independent of a group’s overrid- 24 ing goal. Also, the individual members or factions within a group may not share the leadership’s goals and inter-group competition can influence violence, so motiva- 25 tions alone may not drive tactics and targeting. Furthermore, it can often be diffi- cult to conclusively determine a group’s motivations, as the stated goals may change or not reflect actual tactics. These issues do not indicate that religion is irrelevant in suicide terrorism, only that explanations must be able to account for ambiguity in the goals behind specific attacks. This study, instead, holds with scholars emphasizing the unique nature of religious ideologies, synthesizing these works with social movement theory; it focuses on the mobilization of religion as an ideology, rather than differences in motivations among groups. Several scholars point to the nature of religious beliefs mobilized as a political ideology when explaining the severity of religious violence. That is, the observed effects of religion on terrorism are most likely when groups mobilize the religious tradition as an ideology, especially when they draw on exclusivist or 26 extremist strains of the tradition. A religious ideology—following Moghadam’s definition—is a political program based on a religious tradition in which the group’s interpretation of the religion, rather than ethnic ties or class struggle, is the primary 27 justification for its actions. In the case of Islam, it involves a political platform derived from Islamic symbols, even if the goal is not religious; for example, Hamas aspires to establish a Palestinian state but includes explicit religious justifications for its struggle in its rhetoric. Religious ideology influences behavior at the group level. Few members of groups like al-Qaeda join because of religious fervor, and banal motivations often 28 drive individual terrorists. Yet, religion, like any ideology, can influence a group’s behavior in the absence of individual devotion to its precepts. Ideologies, especially religious ideologies, provide a specific meaning for a group’s actions, enabling group cohesion. Also, the specifics of the ideology can define the means through which groups advance their struggle, with religious terrorist groups characterized by religiously-defined aspirations and dramatic acts of violence to advance their polit- 29 ical aims. The specific type of framing that comes with a religious ideology is what con- nects religious beliefs to the dramatic violence of these groups. As social movement theorists have demonstrated, framing—the manner in which groups justify their 30 actions and interpret their grievances—is a crucial part of a social movement. A group’s framing is an attempt to both win public support for its cause and convince potential supporters of the worthiness of the group’s approach to shared grievances; Downloaded by [Peter Henne] at 07:50 06 December 2011 The Ancient Fire: Religion and Suicide Terrorism 43 the frame will thus influence the group’s tactics and targets. Often, the ideology of a group will be the template from which a group constructs its frames, so groups with 31 a religious ideology will have a distinct type of frame. The rhetoric a group uses to describe its goals and actions—such as appeals to religion or nationalism—can thus have a distinct effect on its behavior. The religious framing of violence by terrorist groups is what leads to the dra- matic effects of religion that scholars have noted. Even if they are motivated by polit- ical grievances, religious terrorist groups base their calculus for success and concerns about public responses on religious standards. The belief in the sacred nature of the attacks and desire to demonstrate the validity of their ideology leads to an emphasis on causing as much destruction as possible from the attack. Likewise, the perception of a global struggle and ‘‘other-ing’’ of ethnic kin who do not abide by the group’s religious beliefs as enemies can lead to a lack of concern for minimizing local fatal- ities. The specific terrorist attacks conducted by religious groups are therefore a form 32 of ‘‘costly signaling’’ that demonstrates commitment to their ideology. Religious framing of suicide attacks does not preclude strategic calculations, although the cal- culations are driven by the nature of the group’s ideology. While this framing influences several aspects of a group’s contention, the most apparent effect is on the severity of its attacks, as noted by the aforementioned scho- lars. Religious groups may have similar goals as other groups and numerous non-religious groups conduct suicide attacks. Once a group decides to use suicide bombings as a tactic, though, the effects of the religious framing will make religious groups distinct from non-religious ones, in line with the theorized effect of religion on violence discussed above. Religious groups will focus on causing as many deaths as possible, and their definition of ethnic kin who do not abide by the group’s religious beliefs as enemies will lead to indiscriminate targeting. In contrast, non-religious groups will either attempt to minimize fatalities to avoid public back- lash or will focus their destructive attacks outside of their own community. Because religious ideology influences groups’ behavior as part of its framing process, the dramatic effects of religion will be evident in groups’ attacks regardless of their goals or organizational nature. This does not completely reject the signifi- cance of organizational characteristics or inter-group competition; it does mean, however, that the apparent influence of religion on suicide terrorism violence cannot be reduced to organizational dynamics. This theory thus corroborates and extends studies on religion and violence that point to the importance of religious ideology, synthesizing them with work on fram- ing in social movement theory and applying them to the specific area of the level of suicide terrorism violence. It agrees with existing studies on the dramatic effects of religion on terrorism, but argues that—when looking at the level of violence from a group’s suicide attacks—religion matters as an ideologically-based frame of con- tention; framing attacks through a religious ideology can lead to dramatic violence even if members’ motivations, pragmatic considerations, and group goals vary. Groups with religious ideologies are thus more likely to cause greater fatalities from suicide terrorist attacks than non-religious groups, an effect that is not reducible to structural factors or group characteristics. Explanations pointing to Islam would expect the suicide attack severity to cor- respond to the presence of Muslims, while non-religious explanations would expect grievances produced by high levels of political repression, ethnic and religious fractionalization, conflict intensity, occupying forces, and low levels of economic Downloaded by [Peter Henne] at 07:50 06 December 2011 44 P. S. Henne development to be most significant. Finally, studies focusing on the motivations of religious groups would expect the distinction between ethnoreligious and fundamen- talist groups to matter more than the broader dichotomy between religious and non-religious groups. Methodology The study uses a compiled dataset with information on suicide attacks and their per- petrators, as well as the socioeconomic, political, and cultural conditions in which the attacks took place. Data on attacks is drawn from a dataset by Kapusta—as part of a project conducted while at the U.S. Army’s School of Advanced Military Studies—which covers all suicide bombings between 1980 and 2006, including the 33 number of dead, location, target, and group (when known). The unit of obser- vation is the suicide attack. While there are several other high-quality datasets avail- able, the Kapusta dataset is particularly useful for this study. The dataset builds on Pape’s definition and draws from multiple sources to ensure reliability. Also it includes numerous suicide attacks that were recorded but failed to result in fatalities, which is useful in minimizing concerns about truncation at zero fatalities. Finally, it allows for comparison over time due to the use of the same methodology for the 34 entire time period. The dependent variable is the number of deaths from an attack, an indicator of a suicide attack’s severity. The number of fatalities provides a measure of an attack’s impact, which can be assumed to flow from the group’s intent to cause harm. Such an approach is in line with the relevant literature, which focuses on the occurrence of a violent conflict, the number of deaths, duration of a conflict, and=or type and 35 number of attacks to measure conflict intensity. As this study is focused on varia- tions in levels of violence between suicide terrorist attacks, the duration of a conflict is not an appropriate dependent variable. Also, while the number of attacks would seem relevant, it captures other aspects of a group, such as how long it has existed. The number of deaths may be influenced by factors besides ideology. It could be influenced by how capable the group is of conducting the attacks, which would be a function of resources or training; it could also be influenced by how effective the tar- get government is at intervening in the attack. Likewise, a group could tend to attack civilian rather than government targets, which may result in more deaths. The study addresses these possible issues through a series of robustness checks and alternate models, which will be discussed below. The dependent variable is continuous but is not uniformly distributed. There are numerous observations on the lower end of the fatalities scale and a few at the high 36 end, so the study will take the natural log plus one of the dependent variable. There are some concerns with censored data due to unobserved failed attacks, but the inclusion of attacks in which there are no fatalities can address this. There are also issues with a lack of independence among observations, as several groups conducted numerous attacks; the methods used will address this issue. Explanatory and Control Variables The explanatory variable, Religious Ideology, is based on the ideological categories in the Terrorist Organization Profiles (TOP), which the National Consortium for the Study of Terrorism and Responses to Terrorism (START) maintains; this variable oD wnloadedb yeP[ tereH nnea ] t:70 eD 60 05 cember2 10 1 The Ancient Fire: Religion and Suicide Terrorism 45 37 indicates whether or not the group has a religious ideology. Some define a conflict as religious through the identity of the combatants, although most combine or replace this factor with a measurement based on the grievances expressed by the combatant 38 groups and how important religion is to the societies in question. Also, several scho- lars explicitly use the ideological profile of terrorist groups as independent or control 39 variables. As this study is focused on the ideology of the groups, it will follow this approach. That is, the explanatory variable is based on the importance of religion in the group’s grievances and the means through which a group justifies its actions, with a 1 for Religious Ideology indicating a religious ideology. Coding derives from TOP ideological categories, and group name is based on that given dataset; groups that are coded as ‘‘religious’’ for their ideology, alone or in conjunction with other motivations, are considered a religious group. Due to possible differences among types of religious groups, two other explanatory variables are coded that are subsets of Religious Ideology. The first, Ethnoreligious, includes groups that have both ‘‘religious’’ and ‘‘nationalist-separatist,’’ listed as their ideology; the second, Funda- 40 mentalist, includes groups whose only motivating ideology is ‘‘religious.’’ Attacks for which the perpetrator is unknown make up nearly half of the obser- vations. Including these attacks with nationalist and leftist groups, though, would be inaccurate. Also, attacks by unknown groups tended to cause fewer fatalities than attacks by both religious and non-religious groups, so including them may bias the results (see Figure 2, below). The study therefore excludes all attacks for which 41 the perpetrator was unknown, in order to focus on the effects of ideology. Several control variables are utilized. In order to test socioeconomic conditions, the study uses the United Nations Development Programme’s (UNDP) Human Development Index (HDI). HDI includes life expectancy, adult literacy rate, edu- cation enrollment, and GDP per capita based on purchasing power parity. The index runs from 0 to 1, with higher human development indicated by a higher numerical 42 score. The dataset used includes data per country in five-year increments; some countries did not include data for specific years, so data were taken from the most recent year’s report available. Iraq and Afghanistan, however, were missing data for several years, which corresponded to the period in which most of the suicide attacks in these countries took place. This study thus uses the most recently available 43 data for these countries, 1998 for Iraq and 1993 for Afghanistan. The study uses regime type as an indicator for political repression. Many measures of regime type, however, include the presence of political violence in their 44 assessments, which could bias the findings of this study. The study therefore uses the ‘‘Democracy and Dictatorship Revisited’’ data developed by Cheibub, Gandhi, and Vreeland (DD), which accounts for this problem, with a 1 indicating the country 45 was classified as a democracy for the year in which an attack took place. The study also makes use of data on ethnic and religious fractionalization. Ethnic Fractionalization (EF) is included through the dataset prepared by Alesina 46 et al. This measures the probability that two individuals selected from a country belong to different ethnic groups, with higher values indicating a higher degree of 47 ethnic fractionalization. Religious fractionalization is addressed through inclusion of data on Religious Minorities; this is a variable developed by Jonathan Fox’s Religion and State Project (RAS), which measures the number of minority religions 48 that account for at least five percent of the population. As sectarian strife often involves tension between several minority religious groups, or a majority and min- ority group, this variable provides an indicator of religious cleavages. Downloaded by [Peter Henne] at 07:50 06 December 2011 46 P. S. Henne The study must also determine whether the presence of Muslims leads to a great- er level of violence in suicide terrorism attacks. This is tested through a variable that indicates if the group is composed of Muslims, regardless of its ideology; this is based on the demographic makeup of the group, using START data (Muslim Group). The effect of conflict levels (Conflict Intensity, CI) on suicide terrorist violence is included through the ‘‘Armed Conflict Dataset,’’ collected by the Uppsala Conflict 49 Data Program (UCDP). This dataset includes all military conflicts between 1946 and 2008. The variable taken from the UCDP dataset is their Intensity measure, which codes conflicts as minor, in which the number of combat deaths is between 25 and 1,000 over the course of a year, or war, with at least 1,000 deaths in any year. These categories are used as a three-level categorical variable in the dataset, with 0 indicating less than 25 deaths, 1 indicating minor armed conflict, and 2 indicating 50 war in the year in which an attack took place. The study also uses several other control variables. A dummy variable, Occupation, is included using Pape’s specification. Pape defines an occupation as a situation in which ‘‘a foreign power has the ability to control local governments inde- pendent of the wishes of the local community,’’ which includes states exerting control 51 over a separatist region. Thus, any country in which military forces—some of which are of a different religious background from the people of the territory— 52 are present in a territory and exert some control over it is coded as occupied. Dummy variables for robustness checks are also used; this includes countries that account for either a great amount of suicide terrorist attacks or deaths—specifically Afghanistan, India=Kashmir, Iraq, Israel=Palestine, Sri Lanka, and the United States—international or transnational attacks, attacks that were successful in caus- 53 ing fatalities, and attacks by an unknown group. There are a few problems with the data that should be noted. First, attacks in occupied or separatist regions were coded for the country in which they resided; the numerous robustness checks should account for any bias that arises. Second, a majority of the attacks were conducted by groups composed of Muslims, and all groups with a religious ideology were Islamic. This would complicate attempts to demonstrate that the effects of religious ideologies differ among Islam, Christianity, and other religions. It will not affect this study, though, because there are numerous Muslim groups with nationalist ideologies and attacks in the non-religious category by groups that are not composed of Muslims, thus ensuring variation in the explanatory variable. Methods This study makes use of a Gaussian GEE model with exchangeable correlation— using robust standard errors—to test the number of deaths, with several robustness 54 checks. The data include fifty-seven groups. A few groups conducted over a hun- dred attacks, while many others conducted less than ten. As a result, the attacks are clustered by terrorist group, with significant clustering among the more active groups. This suggests a lack of independence among observations, which makes ordinary least-squares (OLS) regression inappropriate for this study. GEE models, though, are able to analyze data with such correlation, allowing for more accurate estimations. GEE models include correlation among observations in the model, cor- recting for its effects. The specific GEE model used here is one for continuous data and is population-averaged, which is useful for analyses of variation among groups Downloaded by [Peter Henne] at 07:50 06 December 2011